OpenAI Dissolves Second Safety Team, Removes 'Safely' from Mission in IRS Filing, Restructures as Public Benefit Corporation (2026)

Incident Summary

OpenAI’s governance and safety infrastructure underwent a series of significant changes in early 2026 that, taken together, represent one of the most documented cases of safety culture erosion at a frontier AI company. On February 11, 2026, OpenAI disbanded its Mission Alignment Team — the second dedicated safety team dissolved in two years, following the Superalignment team in 2024.^[1][8]

Concurrently, the word “safely” was removed from OpenAI’s mission statement. The change appeared in an IRS filing related to the company’s corporate restructuring rather than in a public announcement, leading multiple outlets to describe the edit as having been disclosed without fanfare.^[2][7] Under the restructuring plan, OpenAI’s for-profit arm is to become a public benefit corporation (PBC), with the nonprofit retaining control — a structure that differs from a full for-profit conversion.^[7]

Microsoft holds a reported $135 billion stake representing 27% ownership in the for-profit entity,^[4] and SoftBank’s $40 billion investment was reported as conditional on lifting profit caps.^[5] The company has projected $14 billion in losses.^[5]

Evidence from co-founder Greg Brockman’s diary, entered in the Elon Musk trial beginning March 30, included the statement “cannot say we are committed to the nonprofit,” underscoring the documented tension between OpenAI’s original nonprofit mission and its commercial trajectory.^[6]

Key Facts

• Second safety team dissolved: Mission Alignment Team disbanded February 11, 2026 — following Superalignment team dissolution in 2024 (TechCrunch, 2026-02-11)^[1]
• Mission statement change: “Safely” removed from OpenAI’s mission statement in an IRS filing related to restructuring (Fortune, 2026-02-23; The Conversation, 2026-02)^[2][7]
• Corporate restructuring: For-profit arm to become a public benefit corporation; nonprofit retains control (Fortune, 2026-02-23)^[7]
• Microsoft stake: Reported $135 billion (27% ownership) in OpenAI’s for-profit entity (CNBC, 2026-03-23)^[4]
• SoftBank investment: $40 billion, reported as conditional on lifting profit caps (TechCrunch, 2026-03-27)^[5]
• Projected losses: OpenAI projecting $14 billion in losses (per investor documents)^[5]
• Brockman diary: Co-founder’s diary states “cannot say we are committed to the nonprofit” (entered as trial evidence)^[6]
• Legal proceedings: Musk trial beginning March 30, 2026 with nonprofit governance as central issue

Threat Patterns Involved

Primary: Accumulative Risk & Trust Erosion — The sequential dissolution of two safety teams, the removal of “safely” from the mission statement, and corporate restructuring that introduces large-scale commercial investment represent a cumulative pattern. Each action may be individually defensible; taken together, observers and former employees have described them as reflecting a systematic deprioritization of safety at the organization most publicly associated with responsible AI development.

Secondary: Power & Data Concentration — The restructuring, with Microsoft and SoftBank holding a reported $175 billion in combined stake, concentrates governance influence over widely deployed AI systems among a small number of financial actors whose stated incentives are primarily commercial.

Significance

1. Safety structures lack institutional durability — Two safety teams dissolved in two years, combined with the removal of “safely” from the mission statement, suggests that safety structures at OpenAI have not had the institutional permanence to withstand competing commercial pressures
2. Mission evolution with global implications — OpenAI’s transition from a nonprofit AI safety organization toward a commercially-driven entity — documented by its own co-founder’s diary — has implications for global AI governance, as the organization’s original safety-first mission influenced regulatory frameworks and industry norms worldwide
3. Structural tension between safety and returns — The reported $175 billion combined Microsoft-SoftBank stake, with SoftBank’s investment described as conditional on removing profit caps, creates strong structural tension between safety investment and commercial returns that may be difficult to reconcile under the PBC model
4. Industry norm-setting — As one of the most prominent AI companies, OpenAI’s governance decisions influence industry norms; the pattern of safety team dissolutions and mission language changes may signal to the broader AI ecosystem that safety commitments are negotiable under commercial pressure